settled-reach/docs/assets/audio/palette.md

# Audio Palette

The sonic identity of The Settled Reach. Living document — updated as the audio language evolves.

## Core Principle: Audio as Cognitive Architecture

Every sound maps to one of two cognitive sources:

| Source | Sound Type | What it means | Sonic character |
|--------|-----------|---------------|-----------------|
| Biological | Ambient layers, footsteps, fog recognition | "I am a body in a space, and my body knows things" | Organic, warm, natural decay |
| Technological | Cursor hover, implant open, monologue chime | "I am augmented, and my lattice processes things my body can't" | Synthetic, precise, clinical |
| Absence | Dialogue dip, confrontation dip | "I am choosing to focus, and the world continues without my attention" | Reduced ambient, muffled world |

This split maps to D-018's trust model: organic perception is slow but trustworthy, technological perception is fast but its data can be manipulated.

## Two Sonic Families

### Insert-Tech (Synthetic)
- **Character:** Clean waveforms, no reverb (exists in your head, not in space). Digital. Precise. Mathematical.
- **Frequency range:** 800Hz-4kHz primary. Sharp attack, controlled decay.
- **Feel:** "A well-designed touchscreen" — tonal confirmation, not notification beeps.
- **Used for:** Cursor hover, implant open, weapon aim, monologue chimes.
- **Generation method:** Manual synthesis for <200ms sounds. SAO for >200ms.

### Organic (Biological)
- **Character:** Warm, breathy, slightly resonant. Natural attack and decay. Reverberant (exists in the world).
- **Frequency range:** 300Hz-1.5kHz primary. Soft attack, natural tail.
- **Feel:** "The sound you 'hear' internally when something clicks into place."
- **Used for:** Fog recognition, future character-cognition sounds.
- **Generation method:** Stable Audio Open primary — organic texture is what generative audio does best.

## Station Sonic Palette: Functional Warmth

The Settled Reach is a working station. Not military, not luxury — **industrial infrastructure that people made livable.**

### Base Layer (always present)
- **Low-frequency foundation (60-120Hz):** Generator hum, structural resonance. The span gate's mass holding the station together. Always present, barely noticed.
- **Mid-frequency texture (200-800Hz):** Ventilation, distant pipe flow, air cycling systems. Life support you forget about until it stops.
- **High-frequency detail (2-8kHz):** Occasional scanner pings, distant PA crackle, metal-on-metal contact. Sparse. These break through the baseline — something happened.

### Zone Overlays (additive to base)
- **Bar (The Last Shift):** More mid-range (voice murmur 300Hz-3kHz), softer high end (glass filtered, not sharp), faint Meridian music barely melodic — someone left a radio on two rooms away. Social warmth.
- **Corridor:** Emptier mid-range, more prominent low-end (gate hum louder — closer to structure), high-frequency echoes (reverb tail on ventilation whistle). Liminal emptiness.
- **Workplace (Terminal):** Rhythmic mechanical elements (cargo cadence), cleaner scanner pings, distant procedural voices (cargo designations, not conversation). Institutional precision.

### What Silence Sounds Like
Silence on a station isn't silence — it's the base hum with nothing on top. If a zone overlay fades out and you're left with just station base, that should feel **empty**. Not quiet. Empty. Like a room where people were just a moment ago.

## Environmental Neutrality (D-045)

The ambient palette **does not shift with narrative state**. The bar sounds warm whether THE FRIEND is honest or lying. The corridor hum doesn't become ominous when you discover the conspiracy. The world is indifferent to your investigation.

The warmth is the indifference. The bartender's radio doesn't know.

## Audio Bus Architecture

5 player-facing buses, each with independent volume slider:

| Bus | Content | Diegetic source |
|-----|---------|-----------------|
| Music | Future — empty for now | Non-diegetic |
| Ambient | Station hum, zone overlays, global loops | Environment |
| World SFX | NPC footsteps, doors, machinery, conversation murmur | Other entities |
| Player Actions | Player footsteps, combat, weapon sounds | Player character |
| UI Sounds | Cursor, implant, monologue chimes, fog recognition | Neural lattice / cognition |

## Generation Approach

### Stable Audio Open (>200ms, organic character)
- Model: ~1.2B parameters, 44.1kHz stereo output, ~47s ceiling
- Prompting: Describe the SPACE, not the sound. "Industrial space station interior, low frequency machinery hum" not "sci-fi ambient drone in D minor."
- Iteration: Generate 4-6 candidates per asset. Test: solo → stack → fatigue (5+ minutes for loops).
- Post-processing: Trim, normalize (LUFS), EQ to carve frequency space, crossfade loops (3-5s overlap, equal-power).

### Manual Synthesis (<200ms, insert-tech character)
- Tools: Audacity / Python synthesis (sine tone + envelope + harmonic layering)
- The precision IS the design — insert-tech sounds should feel mathematically exact.
- A 50ms cursor tick is a few lines of oscillator + exponential decay.

### Quality Validation
- **Solo test:** Does it sound right alone?
- **Stack test:** Does it sit in the mix with other layers without masking?
- **Fatigue test:** Can you listen for 5+ minutes without jarring repeat? (loops only)
- **Close-your-eyes test (Ozzie):** Does the sound create a mental image / sensation?

## Decision References

- D-018: Three-range sound model (close/medium/long)
- D-038: Audio in v0.1 scope — 8 files via Stable Audio Open
- D-045: Environmental neutrality
- D-048: Neural insert overlay (insert-styled UI aesthetic)
- D-053: Movement stances (acoustic footprint implications)